Electrical Power and Energy Systems 85 (2017) 8796

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Two degrees of freedom dc voltage controller of grid interfaced PV

system with optimized gains
Ravi Nath Tripathi , Tsuyoshi Hanamoto
Graduate School of Life Science and Systems Engineering, Kyushu Institute of Technology, Kitakyushu 808-0196, Japan

a r t i c l e

i n f o

Article history:
Received 14 January 2016
Received in revised form 5 July 2016
Accepted 14 August 2016

Keywords:
Iterative tuning
dc voltage controller
Photovoltaic (PV)
Particle swarm optimization

a b s t r a c t
This paper proposes the application of fictitious reference iterative tuning (FRIT) method to optimize the
gains of dc voltage controller of grid connected photo-voltaic (PV) system. It may be difficult to achieve
good control of dc voltage using conventional PI controller having only one-degree-of-freedom (1-DOF)
due to the trade-off between overshoot (in step response) and disturbance response. In this paper, the
optimal control of dc voltage is proposed with improved disturbance response by implementing 2-DOF
PI controller structure. FRIT method has been programmed in MATLAB based upon the particle swarm
optimization (PSO) algorithm. The fundamental idea related to FRIT method is the extraction of input
and output data, reference model setting and range of controller gains. The performance of dc voltage
control for the optimized 2-DOF PI controller is also compared with the fuzzy logic controller (FLC)
response.
2016 Elsevier Ltd. All rights reserved.

1. Introduction
The use of renewable energy sources have been motivated to
use as distributed energy source for the generation of environmental friendly energy. In the distributed energy source, photovoltaic
(PV), wind, biomass, tidal and geothermal are getting attention in
the last decades. The technological advancement and research
leads to the more feasible and cheaper production of renewable
energy but still it is costlier in comparison to conventional sources
of energy [1,2]. The PV system is extensively popular due to wide
range of power application and received much attention as distributed generating (DG) source in grid connected mode to cater
energy demand problem with inherent propoerty of energy storage
[37]. The PV power generating system as a DG in grid connected
mode having better control operation to operate at maximum
power point (MPP) which will help to evacuate maximum generated power and controlled by grid following power export control
technique [26]. The power electronics interface is essential for
interfacing of PV system as a DG to grid and it can be implemented
using direct/indirect current controlled pulse width modulated
(PWM) voltage source inverters (VSI) [211]. VSI as a power electronics interface supports to implement the grid interfacing and
power conversion [12].

Corresponding author.
E-mail address: ravi1989tripathi@gmail.com (R.N. Tripathi).
http://dx.doi.org/10.1016/j.ijepes.2016.08.006
0142-0615/ 2016 Elsevier Ltd. All rights reserved.

The capacitor dc voltage of VSI in a grid connected PV system

can be regulated using various types of controllers like proportional plus integral type controller (PI/PID), Fuzzy-PI, fuzzy logic
controller (FLC) etc. [215]. Generally, PI/PID controller is extensively used as single input-output system for dc voltage control
and for process control in industrial applications due to ease of
implementation. The conventional PI control structure deals with
the limitation to obtain the optimal response for the system. So,
the 2-DOF PI/PID control structure has been proposed and utilized
under various control schemes for different type of application
with significant advantage over 1-DOF PI control [1620]. The
DOF of a control system defines the number of independent closed
loops in the control configuration [2123]. The 2-DOF control
structure can be design in various control configuration as feedforward, feedback, set point filter etc. [2125].
The PI/PID control structure is designed to obtain the stability
and good control with better transient responses. The performance
of the designed control structure is dependent upon the controller
gains. So, the gains of the controller need to be appropriately tuned
and various tuning methods have been proposed and discussed
like as iterative feedback tuning (IFT), virtual reference feedback
tuning (VRFT) and fictitious reference iterative tuning (FRIT) by
several researchers in [2630]. These direct design approach only
requires the input and output data of the controlled system and
received significant attention. The IFT tuning requires the multiple
experiment data for iterative tuning so as to achieve the desired
optimal response while VRFT required only one-shot experimental

88

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

data. FRIT is also used only one-shot experimental data for the
optimal controller gain tuning. So, these tuning methods helps to
achieve the optimal controller gains without any requirement of
system/plant model information based upon the one-shot closed
loop experimental data. The basic difference between VRFT and
FRIT is that in case of the VRFT minimization of function emphasizes on the input while FRIT emphasizes on the output. So, the
FRIT method may be more advantageous and establishing better
understanding in practical sense. FRIT method implementation
and approach is better for 1-DOF as well as 2-DOF controller in
comparison to VRFT [2933].
FRIT method was modified, studied and analyzed for different
application and optimal performances in [2835]. It can be implemented as both online and offline system for controller tuning. The
FRIT method deals with the disadvantage that it may results in
local minimum for PI/PID controller tuning due to the non-linear
and non-convex problem consideration. So, the FRIT method
should be combined with an optimization technique to mitigate
the local minimum solution for non-linear and non-convex optimization problems.
The DOF based control with optimal tuning methods of controller gains was widely applied and implemented specially in
motor control application and process industry. Sometime, numerical examples are considered to show the effectiveness of DOF controllers and tuning method. This is the first time that the optimal
tuning method and DOF based controller is used for dc voltage control. In this paper, the application of 2-DOF controller is proposed
for the dc voltage regulation of VSI. The data driven tuning method
FRIT is used for optimal tuning of dc voltage controller of grid connected PV system. FRIT is implemented based upon the particle
swarm optimization (PSO) algorithm to prevent the problem of
local minimum solution. The proposed 2-DOF PI control configuration is implemented to mitigate the overshoot in the response of
conventional PI control as well as to achieve the fast transient
response using optimal controller gains. FRIT is used as off-line
tuning method and tune the controller gains by minimizing performance index error function based upon the fictitious reference signal. The performance of the proposed controller is investigated for
step response as wells as under varied ir-radiance as system
response affected due to change in ir-radiance and transient
response.
The paper is organized as follows. In Section 2, the DOF based
controller configuration and design is considered. In Section 3
the optimization problem of FRIT method is considered and implementation in combination with PSO is explained with step by step
procedure considered for 2-DOF PI controller. In Section 4 FLC is
considered and membership function for dc voltage control is
shown. In Section 5 the control algorithm of dc voltage control
with VSI is considered. In Section 6 the performance of controllers
have been discussed and analyzed considering disturbance in reference dc voltage. In Section 7 concluding remarks have been given
for the paper.

2. dc voltage PI controller
The conventional PI control structure in Fig. 1 consists of controller and plant/system transfer function block where, P(s): PI controller (kp + ki/s) and G(s): transfer function of the plant/system.
The 1-DOF PI controller may not achieve very satisfactory response
considering multi-objective view point of the system control and
therefore 2-DOF controller have natural advantage over conventional PI controller [21]. In this paper, the 2-DOF PI control has
been instrumented first time for dc voltage control.
The number of independent closed loops in control structure
decides the DOF of the system control and considered as

Fig. 1. Generalized 1-DOF PI controller configuration.

Fig. 2. Generalized 2-DOF FB type PI controller configuration.

two-input and one output system. The 2-DOF controller designed

and discussed in [16,21] with various configuration stated as
feed-forward type(FF), feedback type (FB), filter type, filter and preceded derivative type and component separated type depending
upon the application and objective of system control. The 2-DOF
feedback compensating (2-DOF FB) controller in Fig. 2 is implemented for dc voltage regulation. The 2-DOF PI control configuration used for dc voltage control consists of proportional gain (K p )
and integral gain (K i ) same as conventional PI controller and additional feedback proportional gain (K pfb ) as a feedback compensation. So, it is having two proportional control gain and one
integral control gain and can be represented as PI-P (2-DOF PI) controller. The reference input signal (r) and controller output signal
(y) are the input signal and the modified output signal (u) output
signal. The closed loop transfer function of conventional PI control
structure in Fig. 1, from reference input r to controlled output y and
from disturbance d to y are given respectively as

Mr1 s

GsPs
1 GsPs

Md1 s

Gs
1 GsPs

The transfer function shown in Eqs. (1) and (2) is having only
one tuning element as Ps for optimal system performance. As it
includes only one tuning element it can-not be tuned independently. The closed loop transfer function of 2-DOF FB control structure in Fig. 2 from reference input r to controlled output y and from
disturbance d to y are given respectively as

Mr2 s

GsPsHs
1 GsHsPs 1

Md2 s

Gs
1 GsPs Hs

Here, the transfer function shown in Eqs. (3) and (4) having two
tuning element as Ps and Hs. It helps to achieve the optimal performance for the system with feedback compensating terms.

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

89

kfb is the feedback controller gain, kpfb is the proportional gain of

feedback controller and initial controller gain for FB is kfb0 . The data
set (u0 ; y0 ) for reference signal computation was obtained by performing one-shot simulation/experiment using P(s, k0 ) and C fb (s,
kfb0 ) with the initial parameter k0 and kpfb0 .
Fig. 3. 2-DOF controller configuration for dc link voltage control of grid connected
PV system.

The controller system is analyzed for the step change in r and

response of the y and termed as reference point (set point)
response. The generalized 2-DOF control configuration shown in
the form of dc voltage control in Fig. 3. The reference dc voltage
(V dc ) is the reference input signal (r) and measure dc voltage
(V dc ) which have to be regulated to reference level is considered
as output for dc voltage control scheme. The 1-DOF and 2-DOF controller response for dc voltage control have been analyzed for the
step change in V dc .
3. Fictitious Reference Iterative Tuning (FRIT)
The performance of the controller is dependent upon gains so
the parameter tuning is one of the significant consideration for better performance. FRIT is data-driven parameter tuning method so it
only requires one shot input and output data. Using this input and
output data it is possible to optimize the controller parameter
without any mathematical model of the system.
3.1. Reference signal generation
The reference signal is generated by using the input and output
data for initial parameter k0 of controller. The k0 was obtained using
trial and error method and represented as [kp0, kint0]T. The one shot
data as: intermediate output u0 as input data and controlled output
y0 as output data. The reference signal generation for conventional
PI control and 2-DOF PI control can be computed as.

3.2. Optimization function and reference model

The optimization of controller parameter can be achieved by
minimizing the optimization function/performance index function.
The optimization function is designed based upon the response
obtained using fictitious reference signal and reference model
and the primary response of the controller for the trial and error
gain.The controller parameters/gains k  kfb have to be tuned and
optimized in such a way that the optimization function should
be minimized. The performance index optimization function is
represented as mathematical equation and based upon system
configuration in Fig. 4. As the FRIT method does not require any
mathematical or dynamical model of the system so a reference
model is to be assumed as transfer function. The generalized transfer function for reference model setting is shown as

Ms

1
ess
Ts 1n

13

where, T is time constant and s is dead time. In this paper, the reference model setting T = 0.0005, s 0 is been used and n = 2. The
controlled output y0 generated using initial parameters is compared
with the output generated using fictitious reference signal and reference model to obtain the optimized controller parameters so that
closed loop response of the system should follow the presumed
ideal response of reference model. The optimization function based
on the initial data set can be defined as

~e y0  Ms~r
I

~e2

14
15

3.1.1. For 1-DOF PI controller

The fictitious reference signal can be computed using Eq. (1)
based upon the system configuration in Fig. 1. The mathematical
evaluation of reference signal for conventional 1-DOF PI controller
is as follows

In the FRIT method, fictitious reference signal generated by Eqs. (6)

and (9) is utilized to generate objective function for optimization
and the optimization function is defined considering the system
configuration in Fig. 4 given as

~r k Ps; k1 u0 y0

~
~ek; kfb y0  y
~
) ek; kfb y0  Ms~r k; kfb

e0 y0

e0 is the error in between r and y0 for the initial response and data
obtained using initial controller gains k0 . The data set (u0 ; y0 ) is
obtained using k0 .
3.1.2. For 2-DOF PI controller
The fictitious reference signal computation can be modified for
2-DOF FB controller in Fig. 2. The reference signal for 2-DOF FB
control configuration can be computed by using transfer function
from r to y in Eq. (3) as

~rk; kfb  y0 Ps; k  Hs; kfb y0 u0

7
1

y0

) ~rk; kfb  y0 u0 y0 Hs; kfb Ps; k

1

) ~rk; kfb u0 Hs; kfb y0 Ps; k

eIk; kfb

2
~ek; kfb

Ps; k kp

k; kfb opt: argmineIk; kfb

Hs; kfb C fb s; k kpfb

19

The optimization algorithm is essential to implement FRIT

method so that local minimum solution can be avoided from the

11
12

18

3.3. PSO based FRIT

10
kint
s

17

So, the objective function for optimization of gain parameters is

to minimize the performance index error function of Eq. (18). It can
be represented as mathematical function of controller gains

where,

k kp ; kint 

16

Fig. 4. System configuration for performance index function.

90

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

FRIT method. The PSO based FRIT (PSO-FRIT) is proposed for the
computation of optimized controller parameters of 2-DOF PI configuration used for dc voltage control. So, an objective function is
required as an optimization problem to achieve optimized controller gains. The optimization method is based upon the behavior
of particle to imitate the potential solution of neighboring particle
which is related to the neighborhood best/local best and its own
achieved solution which is related to the person best solution.
3.3.1. Generalized PSO
PSO method search the possible potential solutions as a swarm
of particles which move in virtual space. So, it is a stochastic
approach and swarm population based search method considering
the behavior of bird flocks. The position and velocity vector is associated with each particle and updated at each iteration. So, the
position of a particle is dependent upon the personal best (pbest )
and neighborhood best which leads to the global best (g best ). Therefore, the particle velocity can be modified as follows

qia1 /qai c1 w1 pbest  xai c2 w2 g best  xai

20

The velocity of particle updated considering following three

terms

considering range constraint for gains based upon the earlier experience of trial and error method.

ki 0 kp;i 0; kint;i 0

ki 0 kp;i 0; kint;i 0; kpfb;i 0

a1

ki

ki qia1

22

The controller gain parameter update in Eq. (22) is adjusted for 2DOF PI controller gains (kp ; kint and kpfb ) as

for PI  P

29
30

The velocity of the particle is also initialized and range constraint is

applied too for proper convergence.

qi 0 qp;i 0; qint;i 0; qpfb;i 0

31

qimax P qi P qimin

32

The pbest and gbest of the particle i and swarm population is also
initialized.
Step 2: The one-shot data (input u0 , output y0 ) is collected using
initial parameters k0 obtained by trial and error method.
The fictitious reference signal generated using the u0 and
y0 will be adjusted and updated for each particle. So, the
performance index is also adjusted related to the reference
signal and error signal for each particle. The updated equation for reference, error and performance are

~r ki Ps; ki 1 u0 y0

33

~eki y0  Ms~rki

34

eIki

21

3.3.2. PSO-FRIT
The general PSO method can be modified for the FRIT method
implementation. The position of particle is defined as the controller gain parameter and update of the velocity represents the
perturbation in controller gains. So, the position update of each
particle in Eq. (21) is modified for controller parameter as follows

28

kimax P ki P kimin

(1) Acceleration coefficient/cognitive component (c1 ), which is

related to the personal best solution (pbest ) of each particle.
(2) Acceleration coefficient/social component (c2 ), which relates
with the best position of particle in whole solution and termed as global best solution (g best ).
(3) An inertia weight (/), which controls the velocity of particle.
The value of w1 and w2 should be decided in between 0 and
1. The position of each particle is modified based upon the
changes in velocity of each particle. The position update
shown as

xia1 xai qia1

for PI

2
~eki

35

~r ki ; kfbi u0 y0 Hs; kfbi Ps; ki 1 y0

36

~eki ; kfbi y0  Ms~r ki ; kfbi

37

eIki

2
~eki ; kfbi

38

Reference, error and performance index equation for 1-DOF PI and

2-DOF PI controller is shown from Eqs. (33)(35) and from Eqs.
(36)(38) respectively. So, the objective function of PSO-FRIT
method for 2-DOF PI is Eq. (39) can be represented as

ki ; kfbi opt: argmineIki ; kfbi

39

Objective function of eq. is evaluated based upon the pbest of each

particle and personal best position is updated for iteration. if

pbesti 6 lbesti

40

lbesti pbesti

41

klocalbest ki

42

then

a1

23

a1

24

and

25

Step 3: The global best position is updated based upon the personal best of each particle. if

a1
kp;i kp;i qp;i
a1
kp;int kp;int qp;int
a1

a1
kpfb;int kpfb;int qpfb;int

The PSO application for FRIT method is illustrated in the following

steps:
Step 1: The number of particles (N) based solution vector of controller parameter is defined as follows

k k1 k2 k3 . . . kj 

26

j 1; 2; 3; . . . ; N

27

The solution vector of controller parameter will be applicable to

the all the three gains (kp ; kint and kpfb ) of 2-DOF PI controller. The
position of each particle with random controller gains is initialized

pbesti 6 g besti

43

g besti pbesti

44

kglobalbest ki

45

then

and

Step 4: The velocity and position of particle has been updated as

according to eq. to eq.

qia1 /qai c1 w1 pbest  kai c2 w2 g best  kai

46

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

The iteration is repeated until the good fitness value is not achieved
considering the constraint of number of iterations as

a amax

47

91

The five fuzzy levels of FLC membership function are: NM

(Negative Medium), NS (Negative Small), ZE (Zero), PS (Positive
Small), PM (Positive Medium). The membership function for output of the FLC is having same number of fuzzy levels and adjusted
as according to desired response.

4. Fuzzy Logic Controller (FLC)

The outer voltage control loop may consist of different type of
controllers along with power compensation control. Fuzzy logic
controller (FLC) is used to regulate the dc link voltage to reference
level and the performance of FLC is compared with the nonoptimized and optimized 1-DOF PI and 2-DOF PI controller
response. The control configuration of FLC in Fig. 5 used for the
dc link voltage control. The error in between the reference value
and output value (E) of dc link voltage and change in error (CE)
are the input quantities to the FLC and output would be the equivalent current quantity for the voltage as the current control algorithm is applied to the system. The inputs of FLC can be
represented as

EK V dc K  V dc K

48

CE Ek  EK  1

49

The fuzzy rule base in Table 1 is implemented for FLC. The

membership function of E is implemented with five fuzzy levels.

5. Control algorithm
The grid interfaced PV system in Fig. 6 consists of LCL filter
and isolation transformer. The power control strategy in Fig. 7 is
implemented based upon the generation of in-phase and
quadrature templates responsible for frequency synchronization
and generation of reference current for active and reactive
power.
5.1. dc link control
The dc voltage of grid connected PV system is regulated by processing error through dedicated controller. The error is defined by
mathematical equation as

EK V dc K  V dc K

50

V dc K

where
and measured voltage V dc K of dc link The output
quantity after the PI control is current as current control method
have been used which is considered in terms of losses. So, the active
power component of current required to compensate the losses and
regulate the dc link voltage to the reference level and can be computation as:
(a) The dedicated 1-DOF PI controller regulates the dc bus voltage to desired reference level & provides the active power
component of current for dc link (id;dc ) to compensate losses.
The value of output current quantity after the PI controller
needed compensate the losses can be computed as

id;dc K id;dc K  1 K pd v de K  v de K  1

Fig. 5. System configuration of FLC for dc voltage control.

K id v de K

Table 1
Controller gains.
E/CE
NM
NS
ZE
PS
PM

NM
NM
NM
NM
NS
ZE

NS
NM
NS
NS
ZE
PS

ZE
NM
NS
ZE
PS
PM

PS
NS
ZE
PS
PM
PM

PM
ZE
PS
PM
PM
PM

51

kp and kint are proportional and integral gain of the PI controller.

Id;dc K and Id;dc K  1 are the loss component of current at Kth
and K  1th sampling.
(b) The dedicated 2-DOF PI controller is replaced by PI-P controller for regulation of the dc bus voltage to desired reference level by feeding losses of VSC. The equation for the
value of output current quantity Eq. (51) is modified as

Fig. 6. Grid interfaced PV system.

92

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

Fig. 7. Power Balance Theory (PBT) based control strategy.

id;dc K id;dc K  1 K pd1 v de K  v de K  1

K id v de K  K pd2 v de K

52

kp ; kint and kpfb are proportional, integral gain and proportional gain
of the PI-P controller. Id;dc K and Id;dc K  1 are the loss component
of current at Kth and K  1th sampling.
5.2. Reference current generation
Unit templates are generated based upon their phase orientation in reference to the grid voltage. The in-phase templates are
generated as [4]

usa

V sa
;
Vt

usb

V sb
;
Vt

usc

V sc
Vt

53

where usa ; usb and usc are the in-phase templates of phase a, b and c
respectively. V t is amplitude of three phase point of common coupling (PCC) voltage and calculated as

V PCC;t

 12

2  2
v sa v 2sb v 2sc
3

54

The equivalent active component of current can be compute by

using the Eq. (55)

iLp

2 pL
3 Vt

55

Therefore, the reference current generated for d component/

active power current of grid current is given as


iLp ip id;dc

56

Active power components of reference grid currents are

Ispa ILp  usa ; Ispb ILp  usb ; Ispc ILp  usc

57

and integral gains obtained through trial and error method.

Extracted data has been utilized to search the optimum gains
through PSO-FRIT programmed in MATLAB-mfile. The response of
conventional PI controller (optimized and non-optimized) is investigated and compared with PI-P controller (optimized and nonoptimized). The dc voltage regulation of grid connected PV system
is further compared between 2-DOF PI-P control and FLC. The transient response during grid coupling of PV system as well as under
the disturbance is investigated and discussed considering simulation results for 1-DOF and 2-DOF PI control. The simulation data
of the system are collected for dc voltage response and plotted
using MS-excel.
In the experimental system, when the switching devices of VSI
were not controlled through PWM signals the dc link voltage of the
VSI system is charged up to a certain voltage level depending upon
the grid voltage rating through anti-parallel diodes of VSI. The
switching of VSI is delayed to show the charging of dc link capacitor and therefore performance of dc voltage regulation to desired
reference level is illustrated when PWM signals are applied to VSI.
The results are discussed and analyzed considering following cases.
6.1. 1-DOF and 2-DOF PI control response
6.1.1. Non-optimized response
The additional proportional controller in 2-DOF PI control is
designed as feedback compensation as shown in Figs. 2 and 3 by
breaking the proportional gain of 1-DOF PI controller into two
parts. The response of the non-optimized traditional PI and PI-P
controller has been illustrated in Fig. 8. The traditional PI controller
is having overshoot with slow response. The objective to design
and implement 2-DOF FB PI control is to minimize the overshoot
by maintaining similar or better response time. In Fig. 8, it is showing improved transient response having no overshoot as well as
settling time is also maintained.

6. Result and discussion

The grid connected PV system is modeled and simulated in the
MATLAB-Simulink environment. The system response for the initial controller gains can be improved using an optimized tuning
method. The input uk and output yk data were collected for the FRIT
method. The uk and yk has been extracted using the proportional

6.1.2. Optimized response

The optimized and non-optimized response of traditional dc
voltage PI control can be observed in Figs. 8 and 9 respectively.
The dc voltage control response is improved for optimized gains
showing fast response and reduced overshoot. Non-optimized
response settled at t = 0.11 s while optimized response settled at

93

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

Fig. 8. Performance of non-optimized dc voltage controller.

Fig. 10. Transient response of dc voltage controller for grid coupling of PV system
and perturbation in reference dc voltage.
Fig. 9. Performance of optimized dc voltage controller.
Table 3
2-DOF controller gains.

Table 2
1-DOF controller gains.
1-DOF PI (trial and error)
1-DOF PI (optmized)

K p 0:3, K i 8
K p 0:77957, K i 29:1989

t = 0.07 s. The peak overshoot is reduced from V dc (overshoot)

= 882.79 V under non-optimized control to 865.94 V under optimized control. The optimized and non-optimized controller gains
for conventional PI controller are given in Table 2.
Optimized response of dc voltage control for 1-DOF and 2-DOF
PI controller is compared as in Fig. 9. The fast response have been
obtained using optimized gains for both the controllers (PI and
PI-P). But, the response of optimized 1-DOF PI is having overshoot
and settled at t = .07 s. The optimized 2-DOF PI response is not
having any overshoot and also the response is even little faster.
PI-P controller response for optimized gains showing improved
transient response with no overshoot as well as fast response
and settling time. The optimized and non-optimized response of
PI-P controller settled at t = 0.065 s and t = 0.1 s respectively.

2-DOF PI (trial and error)

2-DOF PI (optmized)

K p1 0:24, K i 12:5, K p2 0:04

K p1 0:49, K i 28, K p2 0:09

controller. The controller gains for optimized and non-optimized

2-DOF PI controller configuration is given in Table 3.
6.3. Varied Ir-radiance
The performance of the PV system is usually affected due to the
change in ir-radiance level. Therefore, the performance of the control should be investigated under the disturbance caused due
to change in ir-radiance. Fig. 11 shows the dc voltage response
for optimized and non-optimized control of 1-DOF and 2-DOF

6.2. Disturbance in reference dc voltage

The transient response of the controllers has been analyzed
under the disturbance in reference dc voltage. The reference dc
voltage (Vdc ) is altered from 800 V to 840 V (5% of reference dc
voltage) through unit step signal. The response of non-optimized
1-DOF PI and 2-DOF PI is illustrated under the condition of disturbance in reference dc voltage as in Fig. 10. The response of 2-DOF
controller is better in comparison to 1-DOF controller under the
disturbance too. The optimized 2-DOF PI transient response is having no overshoot under disturbance occurred at as well as showing
fast response in comparison to non-optimized 2-DOF and 1-DOF

Fig. 11. dc voltage controller response under disturbance due to ir-radiance.

94

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

controller under the disturbance due to change in ir-radiance. One

of the most critical case of change in ir-radiance is considered to
test the performance of the controller. The ir-radiance is changed
from 1000 W/m2 to 0 W/m2 at t = 0.25 s. and dc voltage get
disturbed. The controller tries to regulate the dc voltage to its

Fig. 12. Comparison of response for FLC and optimized PI controller.

reference level quickly. Optimized 2-DOF PI control is showing best

performance with better dynamic response and having low undershoot under this condition.
6.4. FLC and optimized PI controller response
The FLC transient response of dc voltage control has been illustrated in Fig. 12. The FLC response is having very fast response but
reached to the reference dc voltage at t = 0.075 s. and not having
overshoot in the response. The FLC response have been compare
with the optimized PI controller (2-DOF) response as illustrated
in Fig. 12. The optimized 2-DOF PI controller showing quickest
response as dc voltage is been regulated at t = 0.06 s in comparison
to FLC and optimized traditional PI with time t = 0.075 and 0.07 s
respectively. And, the 2-DOF PI is showing no overshoot in transient response.
In respect of showing transient response for regulation of dc
voltage, the source current performance have been also analyzed
during transient. The source current in Figs. 13 and 14 for the

Fig. 13. Grid current response for 1-DOF PI controller in respect to dc voltage regulation.

Fig. 14. Grid current response for 2-DOF PI controller in respect to dc voltage regulation.

Fig. 15. Grid current response for optimized 2-DOF PI controller in respect to dc voltage regulation.

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796
Table 4
Parameter of electrical utility grid system.
Line impedance (Z g )
Grid voltage (V g )
LCL filter
dc link voltage (V dc )
dc link capacitance (C d )
Base impedance (Z b )
Frequency (f)
Switching frequency (f s )

Rs 0:05 ohm, Ls 1 mH
V LL 415 V
Li 1:4 mH, Lg 1:4 mH, C f 5 lF, Rd 20 ohm
V dc 800 V
C d 1500 lF
Z b 17:22
f = 50 Hz
f s 10 kHz

1-DOF PI and 2-DOF PI-P controller (non-optimized) respectively

during interconnection of PV system to grid. The source current
reaches to steady state at t = 0.115 s for the conventional 1-DOF
PI control and at t = 0.08 s for non-optimized PI-P control. So, the
source current transient response is also improved as the dc voltage regulation response is improved and reaches to steady state
smoothly without overshoot under PI-P control. The source current
for optimized 2-DOF PI-P control is illustrated in Fig. 15 and it is
showing very smooth and fast response in comparison to nonoptimized 1-DOF and 2-DOF controllers. The parameters of the
considered grid connected PV system have been summarized in
Table 4.
7. Conclusion
The 2-DOF PI-P controller configuration is proposed and implemented for dc voltage control of connected PV system. The dc voltage response of conventional PI control and proposed 2-DOF
control is investigated and the performance of proposed controller
found satisfactory as compared to conventional control. The controller gain parameters are optimally tuned using data-driven tuning method. Data driven FRIT method is programmed based upon
PSO algorithm to overcome the drawback of tuning method. The
optimized PI-P controller showing fast and smooth response in
comparison to non-optimized PI-P and traditional PI control during
grid coupling of PV system. As well as, the proposed optimally
tuned controller is also effective under step change in reference
dc voltage (step response) and disturbance due to varied irradiance of PV system. The grid current performance is also investigated during integration of PV system and better response is
achieved corresponding to significantly improved dc voltage
regulation.
References
[1] Yang Bo, Li Wuhua, Zhao Yi, He Xiangning. Design and analysis of a gridconnected photovoltaic power system. IEEE Trans Power Electron 2010;25
(4):9921000.
[2] Azmi SA, Adam GP, Ahmed KH, Finney SJ. Grid interfacing of
multimegawatt of photovoltaic inverters. IEEE Trans Power Electron
2013;28(June):277084.
[3] Al Fadil H, Giri F, Guerrero JM. Grid-connected of photovoltaic module using
nonlinear control. In: IEEE international symposium on power electronics for
distributed generation systems (PEDG), Aalborg, Denmark, June 2528. p.
11924.
[4] Tripathi Ravi Nath, Singh Alka. Design and control of grid interconnected solar
photovoltaic (SPV) system with improvement in power quality. In: Fifth
international conference on power and energy systems (ICPS), Kathmandu,
Nepal, Oct. 2830.
[5] Koutroulis Eftichios, Blaabjerg Frede. Design optimization of transformerless
grid-connected PV inverters including reliability. IEEE Trans Power Electron
2013;28(1):32535.
[6] Katiraei Farid, Iravani Reza, Hatziargyriou Nikos, Dimeas Aris. Microgrid
management control and operation aspect of microgrids. IEEE Power Energy
Mag 2008(May/June).
[7] German solar energy society. Planning and installing photovoltaic systems - a
guide for installers, architects and engineers. Earthscan, 2nd ed.; Reprint
2006.

95

[8] Yang Yongheng, Wang Huai, Zhao Yi, Blaabjerg F. Reactive power injection
strategies for single-phase photovoltaic systems considering grid
requirements. IEEE Trans Indus Appl 2014;50(6):406576.
[9] Terzic B, Majic Goran, Slautez Alojz. Stability analysis of three-phase PWM
converter with LCL filter by means of nonlinear model. Automatika 2010;51
(3):22132.
[10] Tripathi Ravi Nath, Singh Alka, Hanamoto Tsuyoshi. Design and control of
LCL filter interfaced grid connected solar photovotaic (SPV) system using
power balance theory. Int J Electr Power Energy Syst (IJEPES)
2015;69:26472.
[11] Kazmierkowski MP, Malesani L. Current control techniques for three phase
voltage-source PWM converters: a survey. IEEE Trans Indus Electron 1998;45
(5):691703.
[12] Vasquez Juan C, Mastromauro RA, Guerrero Joseph M, Liserre Marco. Voltage
support provided by a droop-controlled multigunctional inverter. IEEE Trans
Indus Electron 2009;56(11):45109.
[13] Louzazni M, Aroudam E. Control and Stabilization of three-phase grid
connected photovoltaics using PID-fuzzy logic. In: IEEE international
conference on intelligent energy and power systems (IEPS). p. 27984.
[14] Gia Minh Thao Nguyen, Uchida K. Control the photovoltaic grid connected
system using fuzzy logic and backstepping approach. In: 9th Asian control
conference (ASCC).
[15] Luo An, Tang Ci, Shuai Z, Tang Jie, Xu Xian Yong, Chen Dong. Fuzzy-PI based
direct output-voltage control strategy for the STATCOM used in utility
distribution systems. IEEE Trans Indus Electron 2009;56:240111.
[16] Miklosovic Robert, Gao Zhiqiang. A robust two-degree-of-freedom control
design technique and its practical application. In: IEEE industry application
conference, Oct. 37.
[17] Lai Yen-Shin, Chang Ye-Then, Kuo Chun-Ting. Robust control of digital
controlled buck converter based upon two-degree-of-freedom controller.
IECON, Glendale, AZ; 2010.
[18] Hanif Moin, Khadikar Vinod, Xiao Weidong, Kirtley Jr James L. Two degrees of
freedom active damping technique for LCL filter-based grid connected PV
systems. IEEE Trans Indus Electron 2014;61:2795803.
[19] Nemati Hosein, Bagheri Peyman. A new approach to tune the two-degree-offreedom (2DOF). In: IEEE international symposium on computer-aided control
system design, Yokohama, Japan, Sep. 810.
[20] Dash P, Saikia Chandra, Sinha N. AGC of a multi-area system using firefly
optimized two degree of freedom PID controller. In: Power and energy systems
conference, Bangalore.
[21] Araki Mituhiko, Taguchi Hidefumi. Two-degree-of-freedom (2-DoF) PID
controller systems. Int J Control Autom Syst 2003;1(4):40111.
[22] Bagheri Peyman, Nemati Hosein. Novel tuning strategy for two-degree-offreedom (2DOF) Pi controllers. In: 18th IFAC world congress, Milano, Italy, Aug.
28-Sep. 02.
[23] Horowitz IM. Synthesis of feedback systems. Academic Press; 1963.
[24] Alfaro VM, Vilanova R, Arrieta O. Robust tuning of two-degree-of-freedom (2DoF) PI/PID based cascaded control systems. J Process Control 2009;19
(10):165870.
[25] Tripathi Ravi N, Hanamoto Tsuyoshi. 2-DOF based novel controller design for
dc voltage regulation of LCL interfaced three phase grid connected solar PV
system. IFEEC, Taipei, Taiwan; 2015.
[26] Hjalmarsson H, Gevers M, Gunnarsson S, Lequin O. Iterative feedback tuning:
theory and applications. IEEE Control Syst Mag 1998;18(4):2641.
[27] Campi MC, Lecchini A, Saversi SM. Virtual reference feedback tuning: a
direct method for design of feedback controllers. Automatica 2002;38
(8):133746.
[28] Lecchini A, Campi MC, Saversi SM. Virtual reference feedback tuning for two
degree of freedom controllers. Int J Adaptive Control Signal Process 2002;16
(5):35571.
[29] Souma S, Kaneko O, Fujii T. A new method of controller parameter tuning
based on input-output data, fictitious reference iterative tuning (frit). In:
Proceedings of IFAC workshop on adaptation and learning on control and
signal processing (ALCOSP).
[30] Tripathi Ravi N, Hanamoto Tsuyoshi. FRIT based optimized PI tuning for dc link
voltage control of grid connected solar PV system. IECON, Yokohama, Japan;
2015.
[31] Kaneko Osamu, Souma Shotaro, Fuji Takao. A fictitious reference iterative
tuning (frit) in the two degree of freedom scheme and its application to
closed loop system identification. Trans Soc Instrum Control Eng 2006;42
(1):1725.
[32] Masuda Shiro. A direct PID gains tuning method for dc motor control using
an input-output data generated by disturbance response. In: IEEE
international conference on control application (CCA), Senver, USA, Sep.
2830.
[33] Kaneko O, Yamashina Y, Yamamoto S. Fictitious reference tuning of the feedforward controller in a two-degree-of-freedom control system. SICE J Control
Measure Syst Integr 2011;4:5562.
[34] Wakasa Y, Tanaka K, Nishimura Y. Online controller tuning via FRIT and
recursive least-squares. In: IFAC conference on advance in PID control, Brescia
(Italy), March 2830.
[35] Tripathi Ravi Nath, Hanamoto Tsuyoshi. Novel controller design for dc
link voltage control of grid connected PV system and optimized PI
response analysis with improved transients. ISGT-Asia, Bangkok, Thailand;
2015.

96

R.N. Tripathi, T. Hanamoto / Electrical Power and Energy Systems 85 (2017) 8796

Ravi Nath Tripathi received his B.Tech (Electrical and Electronics) degree in
2011 from Graphic Era Institute of Technology (GEIT), India and M.Tech (Power
Systems) degree in 2013 from Delhi Technological University (formerly Delhi
College of Engineering), Delhi, India. He is currently working as doctoral student
at the Hanamoto Laboratory, LSSE, Kyushu Institute of Technology, Japan. His
current research is design, modeling and optimization of grid connected PV
system.

Tsuyoshi Hanamoto was born in 1961. He received an M.S degree from Kyushu
Institute of Technology, Japan, in 1986 and then joined Kobe Works of Kobe Steel,
Ltd. In 1990, he joined the Center for Cooperative Research of Kyushu Institute of
Technology. From 1997 to 2000 he was with Department of Electrical Engineering
and since April 2000, he has been with the Graduate School of Life Science and
Systems Engineering, Kyushu Institute of Technology, where he is presently a
Professor. His research interests include motor control and power electronics.